# Is the error rate of IBMQ devices full-scale (% FS) or on percentage of reading (% RD)?

The error rates that I found on the calibration files are given in percentage, but I couldn't find whether it is calculated on full-scale (% FS) or on percentage of reading (% RD). If I do $$20000$$ shots and found $$|00\rangle$$ at $$5000$$ with an error rate of $$1\%$$, should it be $$5000 \pm 200$$ or $$5000 \pm 50$$? Thank you!

The percentage error rates in the calibration files give the average readout error across all qubits. So, in your example, $$20000$$ shots of which $$5000$$ measured $$|00\rangle$$ on a backend with an average qubit error rate of $$1\%$$ would correspond to an estimated uncertainty of $$5000 \pm 50$$. To more accurately quantify measurement uncertainty you'll want to look at the single-qubit error rate for each qubit and the CNOT error rate for each connection, see How to get the CNOT error rate between 2 qubits on a given IBMQ Backend? and Is current information about readout-errors of qubits on IBM quantum devices available on some internet page?